The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill constructs shell commands using untrusted user input (e.g., location names, keywords) without providing sanitization or escaping instructions.
Ingestion points: User-provided strings for locations and keywords are extracted and placed into command templates in SKILL.md (e.g., hpp area small --keyword "<location-name>").
Boundary markers: No delimiters or safety instructions are present to prevent user input from breaking out of the intended shell command context.
Capability inventory: The skill uses the Bash tool to execute CLI commands.
Sanitization: The instructions do not specify any validation or sanitization for the inputs before they are passed to the shell.
[COMMAND_EXECUTION]: The core functionality of the skill relies on executing the hpp CLI via the Bash tool. While the tool is a resource provided by the author (jackchuka), the pattern of interpolating raw user text into shell arguments creates a risk that a malicious user could craft a request (e.g., using backticks or semicolons) to execute unauthorized commands on the host system.

restaurant-search